A method for online evaluation of chromatography column performance and its use

Abstract

The application discloses a method for evaluating chromatographic column performance on line and application thereof, and comprises the following steps: S1, collecting the reading value (conductivity, ultraviolet absorption value, etc.) of the change (rise or fall) of the outlet sensor of the chromatographic column when any different solution (mobile phase) is replaced in the column chromatography process, carrying out normalization processing on the reading value in the change process, and fitting a corresponding normal distribution cumulative distribution function; S2, obtaining column efficiency evaluation parameters based on the cumulative distribution function in the step S1, and the column efficiency evaluation parameters comprise an equal plate height HETP and / or asymmetry Asymmetry. The application aims to provide a novel method (Normal Distribution Transition Analysis, NDTA) for evaluating chromatographic column performance on line based on normal distribution and sensor reading value conversion process, replace traditional offline test method, improve production efficiency, and solve the problems of large calculation error, unstable index, inconsistent dimension, or inability to form effective bridging with traditional methods in the existing TA, DTA, GDTA and other methods.

Description

Technical Field

[0001] This invention relates to the technical field of performance evaluation of chromatography columns, and in particular to a method for online evaluation of chromatography column performance and its application. Background Technology

[0002] Several scholars, including Tina M. Larson, proposed a method for evaluating chromatographic column performance based on the derivative (differential) of conductivity changes during solution displacement (Transition Analysis, TA) in the paper Larson TM, et al. Use of process data to assess chromatographic performance in production-scale protein purification columns[J]. Biotechnol. Prog. 2003, 19:485-492.

[0003] Nathan Mao used this method for online column performance evaluation in chromatography processes in Chinese patent CN109923411B and in US patent Ganguly J, et al., Systems and methods for evaluating chromatography column performance [P]. US8.410,928B2.2013. This method requires calculating the numerical derivative of conductivity. In practical use, due to the significant noise in the conductivity collected by the sensor, noise reduction is usually required. Furthermore, numerical differentiation and noise reduction typically use a fixed number of data points for calculation. When the trend of change is the same but the sampling frequency (different data point densities) is different, the calculation results show significant differences. Commonly used chromatography systems (such as Cytiva's AKTA series) have dynamic sampling frequencies, with different sampling frequencies for different batches of raw data. This leads to unstable calculation results, large fluctuations in indicators, and frequent false positives in practical applications, misleading production process management. In addition, when the conductivity concavity changes during the displacement process, multiple overlapping peaks can occur in the calculated derivative, making it impossible to further calculate the corresponding performance parameters using this method.

[0004] In the literature Cui YY, et al. Using direct transition analysis in chromatography[J]. BioPharm International, 2018, 31(1):34-40, Yuanyuan Cui et al. proposed a method to evaluate the performance of chromatography columns directly based on the change in conductivity during the displacement process (Direct Transition Analysis, DTA). This method establishes an evaluation index for the transition width of the displacement process itself. However, this index differs from the calculation principle of the height equivalent to a theoretical plate (HETP) measured by the traditional method, and the dimensions are also different. It is impossible to establish a mathematical conversion relationship, nor can it be effectively bridged with the traditional method.

[0005] Randolph et al., in their US patent Randolph P. Transition analysis method for chromatography column qualification [P]. US 11,225,516B2.2022, proposed a method called Gamma Distribution Transition Analysis (GDTA) that uses a gamma distribution to fit the conductivity change during the replacement process and calculates chromatography column performance evaluation parameters (HETP and asymmetry) based on the gamma distribution. This method avoids the problems of the aforementioned methods, does not require numerical differentiation calculations, and can establish a certain degree of correlation with the evaluation indicators of traditional methods. However, the calculation of HETP and asymmetry in traditional methods is based on the assumption of a normal distribution (or Gaussian distribution). Since the probability density function (PDF) of the normal distribution itself cannot characterize asymmetry, the asymmetry in traditional methods is calculated numerically. For GDTA, the gamma distribution used can characterize asymmetric peaks. Therefore, its fitting process will be affected by the asymmetry of the peak shape. As a result, when the asymmetry deviates far from 1, the test results of GDTA will deviate from those of traditional methods (including HETP and asymmetry), and the consistency of the calculation results will be poor.

[0006] Tiwari et al., in their paper Tiwari A, et al. Application of advanced machine learning algorithms for anomaly detection and quantitative prediction in protein A chromatography[J]. Journal of Chromatography A. 2022, 1682: 463486, used machine learning to establish the relationship between solution displacement process and chromatography column performance. However, their model is a statistical model, lacking physical constraints, and cannot establish a physical connection with the traditional chromatography column performance evaluation indicators, thus lacking sufficient robustness. Summary of the Invention

[0007] This invention aims to provide a novel online method for evaluating the performance of chromatography columns based on normal distribution and sensor reading conversion process (Normal Distribution Transition Analysis, NDTA), replacing traditional offline testing methods, improving production efficiency, and solving problems such as large calculation errors, unstable indicators, inconsistent dimensions, or inability to form an effective bridge with traditional methods in existing TA, DTA, GDTA and other methods.

[0008] To achieve the above objectives, the present invention provides a method for online evaluation of chromatography column performance and its specific technical solution as follows:

[0009] A method for online evaluation of chromatography column performance includes the following steps:

[0010] S1. Collect the readings (conductivity, UV absorption, etc.) of the sensor at the column outlet when any different solutions (mobile phase) are replaced during column chromatography. Normalize the readings during this process and fit the corresponding cumulative distribution function (G(t) or G′(t)) of the normal distribution.

[0011] S2. Based on the cumulative distribution function in step S1, obtain the column efficiency evaluation parameters, which include equal plate height (HETP) and / or asymmetry.

[0012] In this process, the height of the chromatography column (HETP) is calculated based on the fitting parameters; the asymmetry is calculated using a quadratic fitting method or interpolation method; and the readings in step S1 are conductivity or ultraviolet absorption values.

[0013] Traditional column efficiency measurements employ the pulse injection method, injecting a short conductivity pulse into the mobile phase (e.g., injecting 1 M NaCl into a 150 mM NaCl mobile phase) and observing the peak shape at the outlet. Figure 1 As shown.

[0014] The following provides background information on the theoretical derivation:

[0015] Define the time interval from the injection pulse to the peak value of the effluent as t. r The peak width (half-height width) corresponding to the position halfway up the peak height is W. h The retention time is t;

[0016] Assuming the peak follows a normal distribution, i.e., the sensor reading c satisfies the following equation:

[0017]

[0018] Where μ is the mean of the distribution (μ = t) r ), where σ is the standard deviation of the distribution;

[0019] It is easy to see that c reaches its maximum value when t = μ, denoted as ct. max When c = c max When / 2, t=μ±1.177σ, that is, W h =2.354σ;

[0020] Based on the HETP definition, we can obtain:

[0021]

[0022] Where L is the column height, which is usually expressed in cm in the industry;

[0023] Since the normal distribution cannot characterize asymmetry, asymmetry is usually calculated using the 10% peak height as the standard and actual values; at the 10% peak height, the left peak width is 'a' and the right peak width is 'b'. Based on the definition of asymmetry, we have:

[0024]

[0025] Based on the normal distribution assumption in the tray theory, the function of the outlet peak corresponding to a single pulse in the pulse injection method is derived, denoted as the first function g(t); that is, let the function of the outlet peak corresponding to a single pulse (Δt→0) in the pulse injection method be g(t);

[0026] Based on the normal distribution assumption, that is:

[0027]

[0028] The readings during the change process are normalized to [0,1]. When the pulse is maintained at a constant value from time 0, it is denoted as the second function f(t).

[0029] That is, the sensor readings are normalized to [0,1]. When the pulse is maintained at a constant value from time 0, it is denoted as function f(t), i.e.:

[0030]

[0031] Ignoring the effects of intermolecular forces and concentration gradients on the diffusion process, the cumulative distribution function is macroscopically equal to the convolution of the first function and the second function. The first function is obtained by convolution integration with the second function, which is the derivative of the cumulative distribution function. Thus, the cumulative distribution function G(t) can be obtained from the first function.

[0032] The outlet peak shape is set as G(t), which represents the solution displacement and sensor reading conversion process in a conventional process, such as... Figure 2 As shown; based on the physical meaning of convolution, neglecting the influence of intermolecular forces and concentration gradients on the diffusion process, macroscopically:

[0033]

[0034] We can obtain:

[0035]

[0036] We can obtain:

[0037]

[0038] According to the physical meaning of G(t), G(0)→0, G(+∞)→1, therefore r=0 in the above equation, and we have:

[0039]

[0040] This is the functional relationship that the sensor's reading conversion process must conform to.

[0041] Furthermore, the calculation process for the medium plate height HETP in step S2 is as follows:

[0042] For any process involving solution displacement and sensor reading conversion, let the acquired sensor reading curves be the first dataset. Where (t) i c i ) represents the coordinates of the corresponding point, where the x-coordinate is the retention time and the y-coordinate is the sensor reading. i The data has been normalized to [0, 1] and fitted using the least squares method, i.e.:

[0043]

[0044] The corresponding t for this dataset can be obtained. r and σ, t r Let σ and σ be the peak time interval from the injection pulse to the effluent peak and the standard deviation of the distribution, respectively, and then we can obtain:

[0045]

[0046] The above formula is the calculation formula for HETP (Highest Plate Length) in a chromatography column;

[0047] Where σ is the standard deviation of the distribution, t r Let μ be the time interval between the injected pulse and the peak value at which the pulse emerges, and let μ be the mean of the distribution, μ = t. r L is the column height.

[0048] Since the normal distribution cannot characterize asymmetry, asymmetry can be calculated using the following two methods.

[0049] Furthermore, in step S2, the asymmetry is calculated using a quadratic fitting method as follows:

[0050] Obtain sensor reading curve dataset subset of subset The x-axis of subsets A and B represents retention time, and the y-axis represents sensor readings. The retention time of subset A is less than or equal to the time interval between the injected pulse and the peak value, while the retention time of subset B is greater than the time interval between the injected pulse and the peak value. Subsets A and B are fitted to functions G(t|σ1) and G(t|σ2) respectively using the least squares method, thus yielding the standard deviations σ1 and σ2.

[0051]

[0052]

[0053] For g(t), when the peak height is 10% of the maximum peak height, we have:

[0054]

[0055] We can solve for:

[0056] t 10％ =t r ±2.146σ

[0057] Therefore, the dataset can be obtained. Asymmetry:

[0058]

[0059] Furthermore, in step S2, the asymmetry is calculated using interpolation as follows:

[0060] Although the normal distribution itself cannot characterize asymmetry, the original dataset The data points in the dataset can be directly used for asymmetric calculations, considering G(t). 10％ ):

[0061]

[0062] Based on interpolation, data can be obtained from the dataset. The values ​​of t satisfying c = 0.016 and c = 0.984 are obtained and denoted as t0 and t1 respectively. a t b Then the dataset The asymmetry can be expressed as:

[0063] ;

[0064] The values ​​of c = 0.016 and c = 0.984 are corresponding values ​​based on the traditional pulse injection method for evaluating asymmetry at 10% peak height. These two values ​​are not limited to 0.016 and 0.984. For example, using 5% peak height, 15% peak height, etc., may correspond to different values, which will not be listed here.

[0065] The derivation of G(t) above is based on the process of the sensor reading rising, but it is clear that this process is still valid for the process of the sensor reading falling, in which case the following condition is met:

[0066]

[0067] The corresponding asymmetric readout points are:

[0068]

[0069] By comparing the quadratic fitting method and the interpolation method, the error of the interpolation method can be further confirmed:

[0070] When separately and When the fit is G(t|σ1) and G(t|σ2), it is obtained by solving:

[0071] G(t′ a |σ1)=0.016

[0072] and

[0073] G(t′ b |σ²)=0.984

[0074] We can obtain:

[0075] t′ a=t r -2.146σ1

[0076] and

[0077] t′ b =t r +2.146σ2

[0078] The two schemes have the same computational logic, and the main error is caused by the interpolation of t. a t b The value and t′ obtained by fitting a , t′ b The error between them is usually within an acceptable range.

[0079] Generally speaking, interpolation can yield a more stable solution, but it requires two additional fitting calculations and necessitates judging the goodness of fit (R²). 2 The quadratic fitting method simplifies the calculation process and yields results closer to the original data, but it may be affected by fluctuations or anomalies in the original data, leading to fluctuations and uncertainties in the results (e.g., fluctuations in the original data, resulting in t). a t b (Not unique, etc.)

[0080] Process analysis techniques applied to biopharmaceutical manufacturing processes typically include the following system levels:

[0081] I. Equipment terminals: including but not limited to laboratory-scale chromatography systems, pilot-scale chromatography systems, and production-scale chromatography systems (such as Cytiva's AKTA chromatography system, etc.);

[0082] II. Automatic control systems: such as AKTA's Unicorn control software, or distributed control systems (such as Emerson's DeltaV, etc.);

[0083] 3. Data management system: such as Unicorn software's own database, or other factory information systems (such as OSIsoft's PI system, etc.);

[0084] IV. Data Analysis Platform

[0085] Furthermore, the present invention also provides the application of the above-described method for online evaluation of chromatography column performance in a data analysis platform;

[0086] This invention is mainly applied to data analysis platforms. It is not limited in the selection of equipment terminal types, automatic control systems, data management methods, etc., and can be applied to any combination.

[0087] Furthermore, the present invention also provides the above-mentioned method for online evaluation of chromatography column performance, which can be applied to the process development stage, pilot-scale amplification stage, clinical sample production stage, process validation (PV) stage and commercial production stage in the biopharmaceutical life cycle, and can also be applied to the continuous process verification (CPV) stage.

[0088] Furthermore, the method for online evaluation of chromatography column performance of the present invention can be applied to chromatography systems of various scales, including but not limited to R&D scale systems (e.g., AKTA Pure, AKTA Avant, AKTA PCC 75, etc.), pilot-scale systems (e.g., AKTA Pilot, etc.) and production scale systems (e.g., AKTA Ready, AKTA Process, etc.).

[0089] Furthermore, this method for online evaluation of chromatography column performance can be directly executed in an automatic control system. The execution includes offline execution after data export or online execution through algorithm deployment in a data management system. The automatic control system uses Unicorn control software or DeltaV distributed control system, and the data management system uses Unicorn software's own database or OSIsoft's PI system.

[0090] Furthermore, the present invention is derived using conductivity data as an example, but it can be applied to other data as needed, such as ultraviolet absorption data (UV).

[0091] Furthermore, the method for online evaluation of chromatography column performance of the present invention can be applied to any stage where the performance of chromatography column needs to be evaluated, including but not limited to when the chromatography column is packed, after the column is packed, before the chromatography column is used, when the chromatography column is used in the process, when the chromatography column is used for simulation operation, and after the chromatography column is used.

[0092] Compared with existing technologies (pulse measurement method, TA, DTA, GDTA), the beneficial effects of this invention (NDTA) are:

[0093] I. Compared with the classic pulse measurement column efficiency method, this invention does not require separate offline column efficiency measurement, and can realize real-time online data evaluation, which can save solution volume, operation time and labor costs;

[0094] Second, compared with the column efficiency method (TA) of process analysis based on the conductivity conversion differential of the substitution process, the present invention does not require solving numerical differentials, the calculation results are not affected by the data point density, and the calculation results are more stable and reliable.

[0095] Third, compared with the column efficiency method (DTA) based on conductivity conversion analysis of the substitution process, this invention can directly solve HETP and asymmetry, rather than substitution quantities with different dimensions such as Transition Width, and the measurement results are more comparable to those of the classical pulse method.

[0096] Fourth, compared with the column efficiency method (GDTA) based on gamma distribution fitting of the permutation process conductivity conversion analysis, the present invention is simpler to calculate, uses normal distribution parameters consistent with the pulse measurement method, and is closer to the classical method in characterizing asymmetry, and can better solve the problem of gamma distribution overfitting to asymmetry. Attached Figure Description

[0097] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0098] Figure 1 The image shows a liquid chromatogram, with the vertical axis representing signal intensity and the horizontal axis representing retention time.

[0099] Figure 2 This is a schematic diagram of the solution displacement and sensor reading conversion process;

[0100] Figure 3 This is a flowchart of the present invention;

[0101] Figure 4 The results of measuring asymmetry under the same conditions using four methods—TA, DTA, GDTA, and NDTA—obtained by monitoring the outlet conductivity;

[0102] Figure 5 The results of HETP determination under the same conditions are obtained by four methods (TA, DTA, GDTA, NDTA) derived by monitoring the outlet conductivity.

[0103] Figure 6 The results of determining asymmetry using four methods—TA, DTA, GDTA, and NDTA—under the same conditions were obtained by monitoring the ultraviolet absorbance at the outlet.

[0104] Figure 7 The results of HETP determination under the same conditions were obtained by four methods: TA, DTA, GDTA, and NDTA, which were derived by monitoring the ultraviolet absorption value at the outlet. Detailed Implementation

[0105] To better understand the purpose, structure, and function of this invention, the following detailed description, in conjunction with the accompanying drawings and specific preferred embodiments, provides a method for online evaluation of chromatography column performance and its application.

[0106] Example 1:

[0107] Please see Figure 1-7 This invention provides a technical solution: a method for online evaluation of chromatography column performance, comprising the following steps:

[0108] S1. Collect the readings (conductivity, UV absorption, etc.) of the sensor at the column outlet when any different solutions (mobile phase) are replaced during column chromatography. Normalize the readings during this process and fit the corresponding cumulative distribution function (G(t) or G′(t)) of the normal distribution.

[0109] S2. Based on the cumulative distribution function in step S1, the column efficiency evaluation parameters are obtained, and the column efficiency is measured according to the column efficiency evaluation parameters. The evaluation standard of the column efficiency evaluation parameters is the existing technology, which will not be elaborated here. The column efficiency evaluation parameters include equal plate height HETP and / or asymmetry.

[0110] Obviously, compared with the classic pulse measurement column efficiency method, the present invention does not require separate offline column efficiency measurement. It can realize online evaluation of real-time data based on the changes in the reading of the chromatographic column outlet sensor, which can save solution volume, operation time and labor costs.

[0111] Example 2:

[0112] The following provides background information on the theoretical derivation:

[0113] Based on the normal distribution assumption in the tray theory, a function of the outlet peak corresponding to a single pulse in the pulse injection method is derived, denoted as the first function;

[0114] The sensor readings during the change process are normalized to [0,1]. When the pulse is maintained at a constant value from time 0, it is denoted as the second function.

[0115] Ignoring the effects of intermolecular forces and concentration gradients on the diffusion process, the cumulative distribution function is macroscopically equal to the convolution of the first function and the second function. The first function is obtained by performing a convolution integral operation on the first function and the second function, which is the derivative of the cumulative distribution function. Thus, the cumulative distribution function can be obtained from the first function.

[0116] Specifically, traditional column efficiency measurements employ a pulse injection method, which involves injecting a short conductivity pulse into the mobile phase (e.g., injecting 1 M NaCl into a 150 mM NaCl mobile phase) and observing the peak shape at the outlet. Figure 1 As shown;

[0117] Define the time interval from the injection pulse to the peak value of the effluent as t. r The peak width (half-height width) corresponding to the position halfway up the peak height is W. h The retention time is t;

[0118] Assuming the peak follows a normal distribution, i.e., the sensor reading c satisfies the following equation:

[0119]

[0120] Where μ is the mean of the distribution (μ = t) r ), where σ is the standard deviation of the distribution;

[0121] It is easy to see that c reaches its maximum value when t = μ, denoted as ct. max When c = c max When / 2, t=μ±1.177σ, that is, W h =2.354σ;

[0122] Based on the HETP definition, we can obtain:

[0123]

[0124] Where L is the column height, which is usually expressed in cm in the industry;

[0125] Since the normal distribution cannot characterize asymmetry, asymmetry is usually calculated using the 10% peak height as the standard and actual values; at the 10% peak height, the left peak width is 'a' and the right peak width is 'b'. Based on the definition of asymmetry, we have:

[0126]

[0127] Let g(t) be the function representing the exit peak corresponding to a single pulse (Δt→0) in the pulsed sampling method. The sensor readings are normalized to [0,1]. The pulse is maintained at a constant value from time 0, denoted as function f(t), i.e.:

[0128]

[0129] At this point, the outlet peak shape is set to G(t), which represents the solution displacement and sensor reading conversion process in the conventional process, such as... Figure 2 As shown; based on the physical meaning of convolution, neglecting the influence of intermolecular forces and concentration gradients on the diffusion process, macroscopically:

[0130]

[0131] We can obtain:

[0132]

[0133] Based on the normal distribution assumption, that is:

[0134]

[0135] We can obtain:

[0136]

[0137] According to the physical meaning of G(t), G(0)→0, G(+∞)→1, therefore r=0 in the above equation, and we have:

[0138]

[0139] This is the functional relationship that the sensor's reading conversion process must conform to.

[0140] Example 3:

[0141] Please see Figure 3 Based on Examples 1-2, the calculation process for the medium plate height HETP in step S2 is as follows:

[0142] For any process involving solution displacement and sensor reading conversion, let the collected sensor reading curve dataset be... Where (t) i c i ) represents the coordinates of the corresponding point, where the x-coordinate is the retention time and the y-coordinate is the sensor reading. i The data has been normalized to [0, 1] and fitted using the least squares method, i.e.:

[0143]

[0144] The corresponding t for this dataset can be obtained. r And σ, we can then obtain:

[0145]

[0146] The above formula is the calculation formula for HETP (Highest Plate Length) in a chromatography column;

[0147] As can be seen, the present invention can directly solve for the height of the chromatography column (HETP), instead of using substitute quantities with different dimensions such as Transition Width. The measurement results are more comparable to those of the classical pulse method. Furthermore, the calculation of the height of the chromatography column (HETP) in the present invention does not require solving numerical differentials, and the calculation results are not affected by the density of data points, making the calculation results more stable and reliable.

[0148] Since the normal distribution cannot characterize asymmetry, asymmetry can be calculated using either Example 4 or Example 5.

[0149] Example 4:

[0150] Please see Figure 3 Based on Examples 1-3, step S2 uses a quadratic fitting method to calculate the asymmetry, specifically as follows:

[0151] Obtain sensor reading curve dataset subset of By fitting G(t|σ1) and G(t|σ2) to them respectively using the least squares method, σ1 and σ2 can be obtained:

[0152]

[0153]

[0154] For g(t), when the peak height is 10% of the maximum peak height, we have:

[0155]

[0156] We can solve for:

[0157] t 10％ =t r ±2.146σ

[0158] Therefore, the dataset can be obtained. Asymmetry:

[0159] This method uses quadratic fitting. The proposed method for calculating asymmetry is simpler to perform than existing techniques. It uses normal distribution parameters consistent with the pulse measurement method, and its characterization of asymmetry is closer to that of classical methods. It can better solve the problem of overfitting the gamma distribution to asymmetry.

[0160] Example 5:

[0161] Please see Figure 3 Based on Examples 1-3, step S2, which uses interpolation to calculate the asymmetry, specifically involves:

[0162] Although the normal distribution itself cannot characterize asymmetry, the original dataset The data points in the dataset can be directly used for asymmetric calculations, considering G(t). 10％ ):

[0163]

[0164] Based on interpolation, data can be obtained from the dataset. The values ​​of t satisfying c = 0.016 and c = 0.984 are obtained and denoted as t0 and t1 respectively. a tb Then the dataset The asymmetry can be expressed as:

[0165] ;

[0166] The values ​​of c = 0.016 and c = 0.984 are corresponding values ​​based on the traditional pulse injection method for evaluating asymmetry at 10% peak height. These two values ​​are not limited to 0.016 and 0.984; for example, using 5% peak height, 15% peak height, etc., may correspond to different values, which will not be listed here.

[0167] The derivation of G(t) above is based on the process of the sensor reading rising, but it is clear that this process is still valid for the process of the sensor reading falling, in which case the following condition is met:

[0168]

[0169] The corresponding asymmetric readout points are:

[0170]

[0171] Compared with existing technologies, this method of calculating asymmetry through interpolation does not require solving numerical differentiation, and the calculation results are not affected by the density of data points, making the calculation results more stable and reliable. Moreover, the asymmetry is directly solved, and the measurement results are more comparable to those of the classical pulse method.

[0172] By comparing the quadratic fitting method in Example 4 with the interpolation method in Example 5, the error of the interpolation method in Example 5 can be further confirmed:

[0173] When separately and When the fit is G(t|σ1) and G(t|σ2), it is obtained by solving:

[0174] G(t′ a |σ1)=0.016

[0175] and

[0176] G(t′ b |σ²)=0.984

[0177] We can obtain:

[0178] t′ a =t r -2.146σ1

[0179] and

[0180] t′ b =t r +2.146σ2

[0181] That is, the calculation logic of the two schemes in Example 5 and Example 4 is the same, and the main error is caused by the t obtained through interpolation. a t b The value and t′ obtained by fitting a , t′ b The error between them is usually within an acceptable range.

[0182] Generally speaking, a more stable solution can be obtained through Example 4, but two additional fitting calculations are required, and the goodness of fit (R²) needs to be judged. 2 To confirm the reasonableness of the results; Example 5 simplifies the calculation process, and the calculation results are closer to the original data, but may be affected by fluctuations or anomalies in the original data, leading to fluctuations and uncertainties in the results (e.g., fluctuations in the original data, leading to t). a t b (Not unique, etc.)

[0183] Example 6:

[0184] Process analysis techniques applied to biopharmaceutical manufacturing processes typically include the following system levels:

[0185] I. Equipment terminals: including but not limited to laboratory-scale chromatography systems, pilot-scale chromatography systems, and production-scale chromatography systems (such as Cytiva's AKTA chromatography system, etc.);

[0186] II. Automatic control systems: such as AKTA's Unicorn control software, or distributed control systems (such as Emerson's DeltaV, etc.);

[0187] 3. Data management system: such as Unicorn software's own database, or other factory information systems (such as OSIsoft's PI system, etc.);

[0188] IV. Data Analysis Platform;

[0189] Based on Examples 1-5, the present invention also provides the application of the above-described method for online evaluation of chromatography column performance in a data analysis platform;

[0190] This invention is mainly applied to data analysis platforms. It is not limited in the selection of equipment terminal types, automatic control systems, data management methods, etc., and can be applied to any combination.

[0191] Example 7:

[0192] Based on Examples 1-5, this invention also provides the above-described method for online evaluation of chromatography column performance in the process development stage, pilot-scale amplification stage, clinical sample production stage, process validation (PV) stage, and commercial production stage of the biopharmaceutical life cycle. It can also be applied to the Continuous Process Verification (CPV) stage.

[0193] Example 8:

[0194] Based on Examples 1-5, the method for online evaluation of chromatography column performance of the present invention can be applied to chromatography systems of various scales, including but not limited to R&D scale systems (e.g., AKTA Pure, AKTA Avant, AKTA PCC 75, etc.), pilot-scale systems (e.g., AKTA Pilot, etc.) and production scale systems (e.g., AKTA Ready, AKTA Process, etc.).

[0195] Example 9:

[0196] Based on Examples 1-5, this method for online evaluation of chromatography column performance can be directly executed in an automatic control system. The execution includes offline execution after data export or online execution through the deployment of algorithms in a data management system. The automatic control system uses Unicorn control software or DeltaV distributed control system, and the data management system uses Unicorn software's own database or OSIsoft's PI system.

[0197] Example 10:

[0198] Based on Examples 1-5, the method for online evaluation of chromatography column performance of the present invention can be applied to any stage where the performance of chromatography column needs to be evaluated, including but not limited to when the chromatography column is packed, after the column is packed, before the chromatography column is used, when the chromatography column is used in the process, when the chromatography column is used for simulation operation, and after the chromatography column is used.

[0199] Proof of beneficial effects:

[0200] Proof 1 (based on electrical conductivity)

[0201] Figure 4 and Figure 5 The results of different methods for determining asymmetry and HETP on the same set of three chromatography columns under the same conditions are shown. The mobile phases used in this set of data are 150 mM NaCl and 1 M NaCl, the working linear flow rate is 150 cm / h, and the column efficiency is calculated by monitoring the outlet conductivity.

[0202] For asymmetry, the results of the TA method and the pulse injection method are closest and the overall error is small; the DTA method has low overall sensitivity (slope) and is difficult to effectively capture actual asymmetric variations; the GDTA method has results close to the TA method, but the error is larger, and it deviates significantly in the low asymmetry region, making it unable to effectively capture asymmetric anomalies in this case; the present invention (NDTA method) can effectively capture asymmetric trends under various conditions, and the overall results are consistent with the TA method.

[0203] For HETP (or TransWidth), all four permutation analysis methods exhibited different trends from the pulse injection method under the lowest set of HETP conditions. This may be related to the data differences inherent in permutation analysis methods and requires systematic correction. The TA method, due to the influence of point density on its numerical differential calculations, showed a lower overall result in this example due to the higher data point density. The GDTA method and the results of this invention (NDTA method) were similar, with the GDTA method results being closer to the pulse injection method results and slightly better than the present invention. The DTA method has different dimensions than the other methods, making comparisons on the same scale impossible, but the overall trend was similar.

[0204] Based on the above results, the accuracy scores of each method are summarized in the table below (based on the number and degree of proximity of experimental points to the y=x line, with each experimental point scored from 0.50 to 1.00).

[0205] Method accuracy score TA DTA GDTA NDTA Asymmetry 3.0 0.0 1.5 2.5 HETP 0.5 1.5 2.5 2.0 Total Score 3.5 1.5 4.0 4.5

[0206] It can be seen that, compared with the other three methods, the present invention (NDTA method) has the best overall performance in terms of both asymmetry and HETP.

[0207] Proof 2 (based on UV absorbance)

[0208] Figure 6 and Figure 7 The results of different methods for determining asymmetry and HETP on another set of chromatography columns (three in total) under the same conditions are shown. The mobile phase used in this set of data is pure water and 1% acetone, the working linear flow rate is 150 cm / h, and the column efficiency is calculated by monitoring the outlet ultraviolet absorbance (UV).

[0209] Under these conditions, for asymmetry, the results of GDTA and DTA methods are similar, neither of which can effectively reflect the actual asymmetry (based on pulse injection measurement values) and has low sensitivity. The results of this invention (NDTA method) are closest to the actual values ​​and slightly better than the TA method. For HETP (or TransWidth), the results of TA method are closest to the actual values ​​under these conditions. The performance of this invention (NDTA method) is similar to that of GDTA method, and similar to the results of the previous group. The overall trend of DTA method is similar to that of GDTA method (or NDTA method), but the dimensions are different, so they cannot be directly compared.

[0210] Based on the above results, the accuracy scores of each method are summarized in the table below (based on the number and degree of proximity of experimental points to the y=x line, with each experimental point scored from 0.50 to 1.00).

[0211] Method accuracy score TA DTA GDTA NDTA Asymmetry 2.5 1.0 1.0 3.0 HETP 2.5 0.5 1.0 1.0 Total Score 5.0 1.5 2.0 4.0

[0212] It can be seen that, compared with the other three methods, the present invention (NDTA method) has the best overall performance in terms of asymmetry; the present invention (NDTA method) has the second best overall performance in HETP, only after the TA method. However, considering the instability of TA in calculating HETP (i.e., sensitive to data density and showing huge differences in the two sets of examples based on conductivity or based on ultraviolet absorption), the present invention is more practically applicable than TA.

[0213] It is understood that the present invention has been described through some embodiments, and those skilled in the art will recognize that various changes or equivalent substitutions can be made to these features and embodiments without departing from the spirit and scope of the invention. Furthermore, under the teachings of the present invention, these features and embodiments can be modified to adapt to specific situations and materials without departing from the spirit and scope of the invention. Therefore, the present invention is not limited to the specific embodiments disclosed herein, and all embodiments falling within the scope of the claims of this application are within the protection scope of the present invention.

Claims

1. A method for online evaluation of chromatography column performance, characterized in that, Includes the following steps: S1. Collect the readings of the sensor at the column outlet when different solutions are replaced during column chromatography, normalize the sensor readings during the change process, and fit the corresponding cumulative distribution function of normal distribution. The sensor reading conversion process satisfies the following functional relationship during the rising phase: G(t) is the cumulative distribution function during the rising phase, t is the retention time, tr is the time interval between the injected pulse and the peak value, and σ is the standard deviation. The sensor reading conversion process during the descent phase satisfies the following functional relationship: G'(t) is the cumulative distribution function during the descent phase, t is the retention time, tr is the time interval between the injected pulse and the peak value, and σ is the standard deviation. S2. Based on the cumulative distribution function in step S1, the column efficiency evaluation parameters are obtained. The column efficiency evaluation parameters include equal plate height (HETP) and / or asymmetry. The calculation process for the medium plate height HETP in step S2 is as follows: For any process involving solution replacement and sensor reading conversion, let the collected sensor reading curve dataset be the first dataset, the x-coordinate of the corresponding point be the retention time, the y-coordinate of the corresponding point be the sensor reading, and the sensor reading be normalized to [0,1]. By fitting with the least squares method, the coordinate values ​​of the corresponding points in the first dataset are substituted into the cumulative distribution function of the normal distribution to obtain the peak time interval and standard deviation of the distribution from the injection pulse to the effluent corresponding to the first dataset. Then, the peak time interval and standard deviation of the distribution from the injection pulse to the effluent are substituted into the HETP calculation formula defined in the plate theory to obtain the plate height HETP of the chromatography column. The specific formula for calculating HETP is as follows: Where σ is the standard deviation of the distribution, t r L is the time interval between the injected pulse and the peak value of the effluent; The asymmetry in step S2 is calculated using a quadratic fitting method or an interpolation method. The asymmetry is calculated using a quadratic fitting method as follows: For any process involving solution replacement and sensor reading conversion, subsets A and B of the collected sensor reading curve dataset are taken. The horizontal axis of subsets A and B represents the retention time, and the vertical axis represents the sensor reading. The retention time of subset A is less than or equal to the peak time interval from the injection pulse to the peak value, and the retention time of subset B is greater than the peak time interval from the injection pulse to the peak value. Subsets A and B are fitted to functions A and B respectively using the least squares method based on the cumulative distribution function, thereby obtaining the standard deviations A and B. The asymmetry is the ratio of the standard deviation B to the standard deviation A. The asymmetry is calculated using interpolation as follows: The data points in the sensor reading curve dataset are directly used for asymmetric calculations. Based on interpolation, the time values ​​when the sensor reading equals reading A and reading B are obtained from the dataset, and are denoted as the first time value t. a Second time value t b The asymmetry of the dataset is then expressed as: Among them, t r The time interval between the injected pulse and the peak value of the output pulse.

2. The method for online evaluation of chromatography column performance according to claim 1, characterized in that, The sensor reading in step S1 is either conductivity or ultraviolet absorption value.

3. The method for online evaluation of chromatography column performance according to claim 1, characterized in that, When assessing asymmetry at the 10% peak height, the reading A is 0.016 and the reading B is 0.

984.

4. The method for online evaluation of chromatography column performance according to any one of claims 1-3, characterized in that, The method for online evaluation of chromatography column performance is performed in an automated control system, the execution of which includes offline execution after data export or online execution via a data management system.

5. The application of the online evaluation method for chromatography column performance as described in any one of claims 1-3 in a data analysis platform.

6. The method for online evaluation of chromatography column performance as described in any one of claims 1-3, in the process development stage, pilot-scale amplification stage, clinical sample production stage, process validation stage, commercial production stage, and continuous process validation stage of the biopharmaceutical life cycle.

7. The application of the online evaluation method for chromatography column performance as described in any one of claims 1-3 in R&D-scale systems, pilot-scale systems, and production-scale systems.

Images